Skip to content
View Diya-sudheer's full-sized avatar
0 followers · 2 following

Achievements

Achievement: Pull SharkAchievement: Pair Extraordinaire

Achievements

Achievement: Pull SharkAchievement: Pair Extraordinaire

Highlights

  • Pro

Block or report Diya-sudheer

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Diya-sudheer/README.md

AI, Knowledge Graphs, and Research-Oriented Engineering

I study at Leibniz University Hannover and build intelligent systems across AI agents, knowledge graphs, and data-driven applications. My work combines research ideas with practical software engineering — and every project below ships with tests, a measured evaluation, CI/CD, and a live demo, because claims should be checkable.

Featured projects

Project What it does Proof
medical-knowledge-agent · live demo Role-aware RAG agent: the same medical question answered differently for a patient vs a doctor, grounded in retrieved sources, with honest abstention when nothing relevant exists. FastAPI + Docker. Deterministic eval suite enforced in CI; benchmarked on 1,500 real MedQuAD passages — semantic retrieval lifts Recall@1 0.66 → 0.75 over BM25
document-trust-analyzer · live demo Explainable transparency score for any document: claims vs citations, overclaiming, vague attribution, loaded language — every point itemised. Pure Python, zero dependencies. Detectors measured on a hand-labeled gold set (claim F1 0.77, red-flag F1 0.91), 94% test coverage, coverage-gated CI
better-text-machine · live demo Responsible revision tool for AI-assisted drafts: citation flags, personalization gaps, source notes, trust score, DOCX/PDF import — fully client-side, drafts never leave the browser. Real-browser smoke test in CI; strict CSP; dependency-free vanilla JS

All three demos run the actual project code in your browser (the Python ones via Pyodide/WebAssembly) — no canned responses, no server, no data collection.

Currently building

  • GraphReady — a documents-to-mapping-ready-data pipeline for knowledge-graph construction
  • ANL 2026 automated negotiation agent — opponent modeling and preference concealment
  • A from-scratch knowledge-graph question-answering agent (ReAct loop, tool use, Claude API)

Tech stack

Languages: Python, JavaScript/TypeScript, SQL AI / Data: retrieval & RAG, sentence-transformers, evaluation design, pandas, scikit-learn Semantic systems: RDF, SPARQL, Wikidata/DBpedia, knowledge graphs Engineering: FastAPI, Docker, GitHub Actions, pytest, Playwright, React/Node.js

Research interests

AI agents · knowledge graphs & semantic systems · automated negotiation · evaluation-driven development · trustworthy and transparent AI applications

Popular repositories Loading

  1. skills-copilot-codespaces-vscode skills-copilot-codespaces-vscode Public

    My clone repository

  2. graphready graphready Public

    Agentic document intelligence pipeline: heterogeneous documents to mapping-ready data for Knowledge Graph construction

    Python

  3. medical-knowledge-agent medical-knowledge-agent Public

    A role-aware RAG agent that answers the same medical question differently for patients vs. doctors -- grounded in sources, with a reproducible offline evaluation suite and CI guardrails.

    Python

  4. document-trust-analyzer document-trust-analyzer Public

    Explainable document transparency analyzer -- scores how well text backs its claims with sources and flags overclaiming, vague attribution, and loaded language. Pure-Python CLI, no API key, with a …

    Python

  5. better-text-machine better-text-machine Public

    A responsible document revision tool for source-aware, personal, and transparent writing.

    JavaScript

  6. Diya-sudheer Diya-sudheer Public